📄️ How recommendations work
Recommendation fields, evidence, confidence, and provider-specific behavior.
📄️ work_mem recommendation
Multi-signal work_mem guidance with spill detection, concurrency awareness, and validation.
📄️ max_connections recommendation
Connection limit guidance with memory pressure, workload headroom, and pooler advisory.
📄️ Parallelism tuning
Parallel query and maintenance worker recommendations driven by workload shape and available worker budget.
📄️ WAL and checkpoint tuning
WAL sizing, replica retention guardrails, and checkpoint diagnosis.
📄️ Autovacuum global tuning
cost_limit, max_workers, and cost_delay — recommended together as one decision.